View a Static Warren Demonstration

The WARREN system, named for famed American investor and author Warren Buffet, is an application of the RETSINA multi-agent architecture to the problem of financial portfolio management. A multi-agent system approach is ideal in cases that require the coordination of several component tasks across distributed and flexible networks of information sources and expertise. Whereas a centralized system is susceptible to sytem-wide failures and processing bottlenecks, multi-agent systems are more reliable, especially given the likelihood of individual component failures.

The great advantage of a multi-agent approach is that if one or more agents fail, the system will continue to function reliably. An unresponsive agent is not a problem if agents are designed with redundant capabilities and/or appropriate interagent coordination mechanisms. In instances of component failure, the agent system is designed to reroute tasks to other agents. Task agents may also seek out replacements for unresponsive or failed agents on the network.

The WARREN system builds on current investment practices to deploy a number of different, semi-autonomous software agents. These heterogeneous agents acquire information from and monitor changes to stock reporting databases, interpret stock information, predict the near future of an investment, and track and filter relevant financial news articles. The WARREN system is designed to monitor the ongoing portfolio management process, and thus to function under conditions of extreme uncertainty.

The WARREN agents are derived from the set of reusable software component agents that comprise the RETSINA agent system. This architecture coordinates three kinds of agents: interface, task, and information agents. As the user interfaces graphically and textually with the portfolio manager agent, other agents coordinate tasks, acquire information, and send results, recommendations, and analyses to the user via the interface agent. See the image below for a representation of how these three kinds of agents interact to enable efficient and effective portfolio management:

The interface agent or portfolio manager interacts primarily with the human user as shown in the uppermost part of the diagram, while the set of analysis or task agents coordinate, decompose, and delegate tasks received from the interface agent or from other task agents. Information agents monitor stock and other financial sources. Data culled from the infosphere and stored locally by information agents are sent to one or more task agents upon request, and, following a process of data analysis and integration at the task agent level, are ultimately displayed to the user via the interface agent.

The user's portfolio manager displays a comprehensive summary of the user's portfolio. The interface also allows the user to buy and sell stocks and to request the preparation of a Financial Data Summary or fundamental analysis of the user's stock holdings. The other display available to the user is a price/news graph that dynamically integrates intra-day trading prices and news stories about a stock.

Sub-project: Text Miner

Robotics Institute Project Page

For more information on the WARREN system, see the following publications:



Copyright 2006 - 2012 © Advanced Agent-Robotics Technology Lab - The Robotics Institute - Carnegie Mellon University

Internal Site (Restricted Access).